Fourteen references
Google says Nano Banana 2.1 keeps up to four characters consistent and up to ten objects faithful across up to 14 reference images.
Google · Image model · Released October 6, 2026
Google’s Nano Banana 2.1 takes up to 14 reference images and keeps several characters and objects consistent. On Flowith you can feed it a whole board of references from the canvas.
Prompt: Editorial portrait of a woman in a translucent iridescent raincoat standing in soft rain, a band of prism light across her face, pale blue backdrop, 35mm film grain. Run it
The settings you get in the Flowith composer. Checked on November 14, 2026.
Google says Nano Banana 2.1 keeps up to four characters consistent and up to ten objects faithful across up to 14 reference images.
Banners and tall strips at 8:1 and 1:8, as well as the usual photo and screen ratios.
Explore at 1K, then render the image you keep at 2K or 4K without leaving the canvas.
The button above opens a canvas with the model already selected in the composer.
Attach reference images from the canvas or your library to hold a face, product or style.
Generate, and the image lands on the canvas.
Select the image to edit, upscale, remove the background or vary it, or use it as the start frame of a video.
Each one opens Flowith with the prompt filled in and Nano Banana 2.1 picked. Change anything before you send.
All of these run on the same canvas, so you can try one prompt on several models and keep the best take.
| Model | Sizes | Reference images | Pick it for |
|---|---|---|---|
| Seedream 5.5 | 1K, 1.5K, 2K | Up to 10 | Transparent PNGs, repeated edits |
| GPT Image 2.6 | Up to 3840 px | Yes | Text in images, GIF and SVG output |
| Nano Banana 2.1 | 1K, 2K, 4K | Up to 14 | Many references, extreme ratios |
| Nano Banana Pro | 1K, 2K, 4K | Yes | 4K detail |
Text, image and video models side by side. Compare them on the same prompt, pick one per shot, or let the agent choose.
Write it, picture it, film it. One sentence is enough to begin.